The literature contains two alternative hypotheses for the mechanism of dehydration of fructose to 5-(hydroxymethyl)-2-furaldehyde (HMF), namely (1) a sequence of reactions commencing with and retaining the fructofuranose ring intact, and (2) a succession of reactions proceeding mainly via open-chain intermediates. The existing evidence for hypotheses (1) and (2) is reviewed and found to favor (1). The major products from fructose in water at 250 degrees, (with and without acid catalysis) have been investigated on a time-resolved basis and analysis of the results was found to confirm the first hypothesis. A necessary fructofuranosyl-cation intermediate in this hypothesis is produced directly by the hydrolysis of sucrose, and reacts to produce HMF in high yields.